由於硬體的進步，越來越多的演算法得以實踐，並且獲得豐碩的成果，例如Google DeepMind所開發的AlphaGO是第一個稱霸圍棋界的電腦圍棋程式。然而AlphaGo應用了非常巨量的人類知識來輔助學習，如:職業棋士對局的棋譜、人類標示的棋盤資訊…等等。 Google DeepMind提出了不需要人類知識，僅依靠MCTS方法訓練的Alpha Zero，可惜的是並沒有開源，故本論文將依Google DeepMind提出論文揭露之訊息嘗試實作Alpha Zero方法於迷你將棋遊戲。

Due to the advancement of hardware, more and more algorithms have been implemented and have achieved fruitful results. For example, AlphaGO developed by Google DeepMind is the first computer Go program to dominate the world of chess. However, AlphaGo uses a very large amount of human knowledge to assist in learning, such as: the chess game of professional chess players, the chessboard information marked by humans, and so on. Google DeepMind proposes Alpha Zero, which does not require human knowledge and only relies on the MCTS method. Unfortunately, there is no open source. Therefore, this paper will try to implement the Alpha Zero method in the mini chess game according to the information disclosed by Google DeepMind.

摘 要I
AbstractII
致 謝III
目 錄IIV
圖目錄 VI
表目錄VII
第一章 緒 論 1
11 研究背景1
12 迷你將棋簡介 1
13 研究動機與目的 3
14 論文概述3
第二章 文獻探討 4
21 蒙地卡羅樹搜尋(Monte-Carlo Tree Search,MCTS)4
22 AlphaZero 中的 MCTS6
23 卷積類神經網路(convolutional neural network, CNN)8
24 深度殘差網路（Deep residual network, ResNet）9
25 AlphaZero 中的神經網路10
26 AlphaZero 中的特徵提取 11
第三章 研究方法 12
31 棋盤特徵提取 12
32 神經網路模型設計 13
33 神經網路模型訓練 14
第四章 系統開發 15
41 主程式設計 15
42 圖形化介面設計 16
第五章 實驗結果 17
51 實驗環境17
52 模型訓練結果 18
53 各模擬次數對局結果比較 19
第六章 結論與未來展望 20
63 結論20
62 未來展望20
參考文獻21

[1] D. Silver, A. Huang, C.J. Maddison, et al, “Mastering the game of Go with deep
neural networks and tree search”, Nature Vol. 529, 2016.
[2] David Silver, Julian Schrittwieser, Karen Simonyan, Ioannis Antonoglou, Aja Huang, Arthur Guez, Thomas Hubert, Lucas Baker, Matthew Lai, Adrian Bolton, Yutian Chen, Timothy Lillicrap, Fan Hui, Laurent Sifre, George van den Driessche, Thore Graepel, and Demis Hassabis. Mastering the game of go without human knowledge. Nature, 550:354– 359, 2017.
[3] Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., et al. Mastering chess and Shogi by self-play with a general reinforcement learning algorithm. arXiv preprint arXiv:1712.01815, 2017
[4] Wikipedia, MiniShogi
, https://en.wikipedia.org/wiki/Minishogi, access@May.11,2019
[5] Andrej Karpathy, “CS231n: Convolutional Neural Networks for Visual
Recognition”, http://cs231n.github.io/, 2016.
[6] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2015. Deep residual learning for image recognition. arXiv preprint arXiv:1512.03385.